What's new
What's new

Drawing/Model discrepancies disrupting our schedule

Of manufacturing and the expectations of dimensional and geometric qualities that can be achieved. I have 1980s manufacturing magazines that brag about how "CNC technology can now maintain tolerances better than 0.010"". Forgive me if that's not impressive anymore. Technology moved on and we got worse.

Here's another example: supply a drawing and 3D model to manufacturing. They inexplicably screw up the location of basic mounting holes. Walk over and talk to them. Eventually get out of them that they were at the end of their shift so they broke down the machine and went home without the holes. They did them the next day. Stupid but, okay. Why didn't they indicate it back into the machine? "well, you didn't specify a tolerance and I can hit the +/- 0.020 in the title block without checking."

Was that lazy or stupid? The minute someone puts 3 place GD&T requirements, it gets sent off to the CMM and spends more time in inspection than they spent trying to figure out the crappiest way they could make it and squeak by.

Another: try to give manufacturing some breathing room by specifying wider flatness and parallelism zones on what should be a datum surface. The intent is to accept something that maybe twists during manufacturing but, will acceptably flex enough on assembly to not matter. There's maybe a center bored hole that has to be some squareness to the surface.

What we get back is something where they took advantage of the wide tolerance to hit the raw stock with a DA sander (because the zone was wide enough to include everything on that surface). Then they did a shitty bore job on a knee mill and barely made squareness because "your drawing allowed that and it meets the specification." Yes, but no. I didn't want a piece of crap to deliver to our customer.

These would be examples of not meeting "best shop practice". I agree that's an ambiguous note but, I don't know any other way to communicate: don't produce crap. It's the 21st century. Doing this in a CNC isn't a hardship.

Modern CNC equipment does not provide anything new or better in terms of geometric precision. Not in any meaningful way at least. Good quality 5 axis machines make it easier, but are not inherently more accurate than say an optical dual rotary table.

I still think you're missing the point. Maybe we work with different types of parts. It has nothing to do with CNC or how manufacturing capabilities have advanced. If a part has some critical multi-directional perpendicularity requirement between discreet features or a finite and quantifiable maximum cylindrical error of a round feature, that must be met or the part will not function, how do you define that on a drawing unambiguously? How does the machinist know that these things matter, if it cannot be assumed that the requirements will be met, regardless of how advanced the equipment is that they have to perform the work, unless it is specified with standardised methodology in a way that the machinist can understand and take whatever necessary precautions to ensure it?

I know that's verbose, but it's important. GD&T fills a very necessary role, and until someone can actually replace it with something better, it's what we've got.

I don't like GD&T any more than the next guy, but I understand why it's necessary and I personally don't have any objectively better ideas on how to perform its function.
 
If a part has some critical multi-directional perpendicularity requirement between discreet features or a finite and quantifiable maximum cylindrical error of a round feature, that must be met or the part will not function, how do you define that on a drawing unambiguously?
Yes, for that particular application, you would use GD&T. Let's use a two-stroke motorcycle cylinder for your example. Yes, I should define the requirements for the orientation of the center bore. The drawing might include the bottom surface and center bore as datums, some squareness tolerances. And that is all a hard-maybe. The fact that it's going to be bored and then honed to some finish requirement is all that should matter. Those processes will inherently create the feature necessary.

Beyond that, the designer should not need GD&T to describe the fastener hole locations, the heights and widths of any of the transfer ports, the shape of maybe a reed cage hole, the clocking of the ports to the finished product, or anything else. Common shop practice should put all of those features well within what they need to be.

Again: shouldn't need to specify no rotting food when you go to a restaurant. It's implied.
 
Yes, for that particular application, you would use GD&T. Let's use a two-stroke motorcycle cylinder for your example. Yes, I should define the requirements for the orientation of the center bore. The drawing might include the bottom surface and center bore as datums, some squareness tolerances. And that is all a hard-maybe. The fact that it's going to be bored and then honed to some finish requirement is all that should matter. Those processes will inherently create the feature necessary.

Beyond that, the designer should not need GD&T to describe the fastener hole locations, the heights and widths of any of the transfer ports, the shape of maybe a reed cage hole, the clocking of the ports to the finished product, or anything else. Common shop practice should put all of those features well within what they need to be.

Again: shouldn't need to specify no rotting food when you go to a restaurant. It's implied.
You take way more for granted than I do. Your non sequitur restaurant example being the exception.
 
You take way more for granted than I do. Your non sequitur restaurant example being the exception.
Let's flip this: using the part in post 18, nothing but the 3D model and a drawing with only thread / ream callouts (if any)? Positions, parallelism and squareness are implied by the shapes and confirmed in the 3D. For scale, lets assume it's a foot wide. Material is 6061.


What parallelism, position and squareness could you guarantee without inspection? Can you hold 1/4" and 1 degree? How about half that? A tenth? What precision could you deliver where you're so confident that you wouldn't even take it to inspection?
 
What precision could you deliver where you're so confident that you wouldn't even take it to inspection?
Let's just say we work differently. I never have the luxury of working to 1 degree or 1/4 inch. Nor do I have to work to a tenth. A few tenths sometimes. It's not worth scrapping 1000 pcs because I didn't check something. Shit happens when you least expect it, and it's never good. And there is that pesky 100% dimensional inspection thing to deal with.
 
No, no, not a tenth of a thousandth. A tenth of the 1/4" (0.025") and tenth of a degree. Pretend I'm a customer and I need parts to be as cheap as possible. I'm going to waive inspection. How tight can you guarantee?
 
No, no, not a tenth of a thousandth. A tenth of the 1/4" (0.025") and tenth of a degree. Pretend I'm a customer and I need parts to be as cheap as possible. I'm going to waive inspection. How tight can you guarantee?
Sorry just not in my DNA not to check stuff. Plus it's a moot point. That's not my type of customer.

Best of luck to you.
 
Not in my experience. I had two mating parts with twenty tapered protrusions (basically conical dowels), all the way around a 24" diamond-shaped opening in aluminum. The towers were ball-milled on a "supposedly" shitty Haas VF-5XT. The mating bores were also ball-milled on the same machine.

The parts mated perfectly in one orientation and were out maybe 0.0005" if spun 180 degrees. It was enough to jam the assembly in the second orientation but just slightly.

It was the first time we discovered the machine wasn't completely square. A 0.257" hole and 1/4" fasteners would have been so sloppy we'd have never known. It would have worked with 0.250" holes if that had been the design. The experts said we couldn't do it and we did it on the first try. That Haas did lots of things we were told wouldn't work.
3 points define a plane
why are there 24 tapered dowels?
trying to prove a part cannot be made?
Bad design
half a thou over 24 inches, this is not VMC tolerances
 
3 points define a plane
why are there 24 tapered dowels?
trying to prove a part cannot be made?
Bad design
half a thou over 24 inches, this is not VMC tolerances

Because the dowels took the place of fasteners and had to carry shear load through the panel. They had to be zero clearance or as close to it as practical. Yeah, people said bad design and wasn't possible. It exists. NDA and all. Can't share pictures.

Made multiple versions of it, spanning years. Same Haas. You're telling me what you believe. I'm telling you what I've done.
 
You do excellent work. 1* F in aluminum over 24" equals .0003" change. Hope you wore gloves!
Both parts start and live in the same environment. Not a problem.

I'm not telling anyone they can't do this. I'm saying you probably are already doing it and with the slightest of care in the process, it could be done repeatably and reliably.
 
I laugh at your discrepancies. Hole tolerance/treatment on print is not contradictory to model. Seems helpful almost.
Steel is moving to generated models and prints (tekla/sds). I will have holes on the model, not on print, and a different size beam all together on the Rivett model. No matter what you go off it will be wrong- Murphy’s law. Also means if rivett and model do not match then there is a problem from the get go.
I have stairs now called out on drawing with beams for stringers, modeled as flat bar, drawn as tubes. I know drawing material is right, but it has imaginary dimensions.
This is everyday thing. Was bad for a while, then ok, latest tekla update is making me loose hair. They have moved to truncates models (20’ beam modeled at 7’ or ‘18’ feet to make neater print). Machines do not read prints. Our models are all dstv, so straight to machine.

Ps I hate generated drawings....
 
We ran into issues like this too many times from not setting rules to the customer, someone needs to place responsibility to the customer, and to the machine shop.
I now tell customers parts will be programmed from "your" solid model, and we are "only" responsible for the specifications on your prints, period.

Another thing that helps us is that we make setup prints and inspection sheets from their solid model, that only has the critical dimensions needed for that OP, and when making it the dimensions must match their drawing,
since that is all we are responsible for, not their model.

Just had a customer yesterday, brought in a part to be modified, and a solid model, nope!
I can use your solid model, but am only responsible for the specifications on the print, so they made one.

I say the same phrase to all customers, "a machinist makes parts from your/mine models to the specifications/tolerances on the print".

Some customers have neither, I make the model, and the print, but they have to sign off on the print as to accept responsibility to its accuracy.
 
Modern CNC equipment does not provide anything new or better in terms of geometric precision. Not in any meaningful way at least. Good quality 5 axis machines make it easier, but are not inherently more accurate than say an optical dual rotary table.

I still think you're missing the point. Maybe we work with different types of parts. It has nothing to do with CNC or how manufacturing capabilities have advanced. If a part has some critical multi-directional perpendicularity requirement between discreet features or a finite and quantifiable maximum cylindrical error of a round feature, that must be met or the part will not function, how do you define that on a drawing unambiguously? How does the machinist know that these things matter, if it cannot be assumed that the requirements will be met, regardless of how advanced the equipment is that they have to perform the work, unless it is specified with standardised methodology in a way that the machinist can understand and take whatever necessary precautions to ensure it?

I know that's verbose, but it's important. GD&T fills a very necessary role, and until someone can actually replace it with something better, it's what we've got.

I don't like GD&T any more than the next guy, but I understand why it's necessary and I personally don't have any objectively better ideas on how to perform its function.
You know what's coming next, due to this thread, standardized imbedded GD&T in the CAD model, transferable through file types, there by avoiding prints, :bawling:
 
You know what's coming next, due to this thread, standardized imbedded GD&T in the CAD model, transferable through file types, there by avoiding prints, :bawling:
I'm seeing videos going back at least five years. This isn't coming next. It's been here. It's not widely used but, it's here. This is the next thing people will get to push back on.

 
Wow. This blew up in a day. Don Quixote and Finegrain get it. Seymore seems to either want a fight or just not get it. I was trying to be polite, but I forget this is PM...

We make injection molded parts, mainly. An occasional metal part, and a LOT of machined prototypes and fixtures. And it is ALL done in 3D CAD. The print, FOR US, as noted by several is an INSPECTION drawing which lists a very small number of important dimensions for part release based on what is important, and what may actually move during production. If I gave you one of our prints only and said "make this" you could not. There is simply not enough information on the print to make the part. That is 100% intentional.

But every plastic part we make requires a mold. We spend lots of time with the mold maker making sure what they see is what we see and getting it right. If there is something like a metal insert needed, we will tell them and work through it. We won't wait until the first part is molded and say "where the heck is the insert?" That is dumb beyond belief. And mold makers are machinists, and they seem to be fine with all this...

Now for prototypes and fixtures, it is a mix. If it is going into production we will make a print, again with key dimensions to ensure the fixture will work. But prototypes? It is very rare to make a print for us as we send a STEP or other CAD file and they make the part. But geesus - not in a vacuum! We talk to the guys making it. And say we need this to be critical, and this isn't important, and whatever. Anything that can't be communicated by the model is worked out directly, or we send a sketch or whatever to make it clear. I like my suppliers and want good parts, so I work with them to that end. Some of you guys must have the worst idiots for customers on the planet based on what you are saying here. Do mistakes happen? Hell yes, but we work through them.

But if you want to work for us, you are getting a 3D model, so if that is unacceptable, then our business models do not match, and we shall go our own ways.
 
But if you want to work for us, you are getting a 3D model, so if that is unacceptable, then our business models do not match, and we shall go our own ways.
Pretty sad you did not get anything some of us were saying.

In fact, you are doubling down on your own contradiction:

Anything that can't be communicated by the model is worked out directly, or we send a sketch or whatever to make it clear.

Please explain to me how is that to be interpreted?
To me, it means that all information is to be gathered from the model, but information contained on the attached drawing override any and all model data.

This post is the best indication of what I am talking about:
Sometimes we get a drawing that does tell us.

NOTE: 6

REDUCED INFORMATION DRAWING. ALL DATA SPECIFIED ON THIS DOCUMENT HAS
PRECEDENCE OVER THE SOLID MODEL FILE
. ALL UNSPECIFIED DATA IS CONTROLLED
BY THE SOLID MODEL FILE.

Pretty clear that the drawing has control over the solid model.
Seems to me that it is you who want to argue...
 
Nah. We work very well with our suppliers. You seem to be the one who wants to argue, no matter what you say. Believe what you want.
 
You know what's coming next, due to this thread, standardized imbedded GD&T in the CAD model, transferable through file types, there by avoiding prints, :bawling:
I'm seeing videos going back at least five years. This isn't coming next. It's been here. It's not widely used but, it's here. This is the next thing people will get to push back on.

Yes, internally I've been using PMI/MBD for years to communicate between all of us inside these four walls. MUCH more convenient than drafting a drawing.

It's been at least a decade+ since it was introduced. I used it heavily in NX10 which came out in 2014.

Unfortunately, there is still no standard interchange format for it which makes it useless unless everyone involved is using the same cadcam system. I believe the most recent iteration of STEP laid some foundation for it, but not usably implemented yet.
 
I think you need both (most of the time), and there ought not be a "Master". They should not conflict, and if they do it's an issue that needs to be cleared up.

I missed this post earlier.

The problem is that they effectively always conflict on some level, and there are good reasons for why they do. The whole reason that I make my own solid models is that I can model them to exactly reflect how I want the physical part to come out, which makes it simpler and more reliable for programming and also usually makes it easier to use the model for inspection.

Most engineers will not model tolerances into a part - they'll model nominal sizes or they'll model some combination of minimum or maximum material to ensure assembly clearances etc. I do this myself when I'm designing a part rather than making a part. It just makes sense at the design stage, but the side effect is that the "design" model often cannot be considered a "master" model for manufacture.
 








 
Back
Top